[HDDS-2198] SCM should not consider containers in CLOSING state to come out of safemode

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 0.5.0
• Component/s: SCM
• Labels: pull-request-available
• Target Version/s: 0.5.0
• Assignee: Nandakumar
• Reporter: Nilotpal Nandi
• Created: 27/Sep/19 14:26
• Updated: 24/Feb/20 20:55
• Resolved: 04/Oct/19 02:52

Description

There are cases where SCM can be stuck in safemode for ever if it considers containers in CLOSING state for coming out of safemode

• If there are 5 containers in OPEN state inside SCM
• Out of 5, 3 containers are created in datanodes by the client.
• 2 containers are yet to be created in datanodes
• Due to some pipeline issue, pipeline close action is sent.
• All 5 container's state are changed from OPEN to CLOSING in SCM.
• Eventually , 3 container's state moves from CLOSING to CLOSED in SCM as the datanodes closes those containers.
• 2 of the containers are still in CLOSING state.
• SCM is restarted.
• SCM will never gets container reports for the containers which were in CLOSING state as those containers were never created in datanodes.
• SCM will remain in safemode.

Issue Links

• links to: GitHub Pull Request #1540